A few problems with that article:
1) The GTX 570 and 6950 are not comparable cards. The higher performing GTX 570's will always be more bottlenecked by the CPU than the 6950's. To gain a more accurate picture, they would need to test 6970's in Crossfire.
2) Only 6 benchmarks does not give a complete picture. Add in that 3 of the 6 benchmarks notoriously favor AMD cards (F1, 3D Mark 11, Aliens vs. Predator), then you further skew the results.
3) The chart on Load Temps is indicative of the questionable nature of the review. 62c at load on GTX 570's in 3-way SLI? The 6950 in 3-way Crossfire only tops out at 58c? These results are quite astonishing, and are simply not supported by other review websites, or common sense.
4) Overall, this article just doesn't seem to match up with other reviews you will find around the web.
Here's one review that does a much better job, for example, 6970 Crossfire vs. GTX 570 SLI. It uses multiple benchmarks including those that favor each brand and some that are neutral. These benchmarks average out to 84% scaling for the 6970 in Crossfire, and 84% for the GTX 570 in SLI at 2560x1600, AF 16X, AA 8X(4X) (top chart on the right):http://www.xbitlabs.com/articles/video/display/radeon-hd6970-crossfirex_6.html#sect0